The sign reversal problem in structural decomposition analysis

A structural decomposition analysis (SDA) based on the input-output model disaggregates fluctuations in the total factor budget into shifts in its determinants. The essence of SDA is its ability to quantify the critical factors that contribute to changes in phenomena. However, it is well-known that various uncertainties are manifest in input-output datasets and SDA results may be vulnerable to substantive biases including erroneous sign reversals. This study employs Monte Carlo simulations and investigates this sign reversal problem. The simulations reveal instability in the decomposition results, particularly the effects of the intensity term and the economic structure term. In contrast, the decomposition effect of the final demand term is relatively insusceptible in this regard.